False Failure in Flexural Fatigue Tests

F. Prez-Jimnez Technical University of Catalonia, Barcelona, Spain

R. Mir, A. Martnez, R. Botella

Technical University of Catalonia, Barcelona, Spain

O. Reyes
Nueva Granada Military University, Bogot, Colombia

G. Valds
Universidad de La Frontera, Temuco, Chile

ABSTRACT: Flexural fatigue tests are typically run under displacement or in a straincontrolled mode. In these tests, either the oscillatory displacement amplitude or strain amplitude applied to the bottom of the specimen is kept constant. The evolution of loading required to cause fatigue is then measured. Load amplitude decreases with the number of cycles, and the specimen is considered to have failed when the load is half its initial value. This failure criterion may be erroneous when non-fragile fracture mixtures prepared with high bitumen contents or modified binders are tested. In these cases, mixtures exhibit a visco-plastic behaviour and increasingly less stress is necessary to cause strain without cracking. Mixtures are hardly deteriorated when the fatigue failure is determined, and may be subjected to a larger number of load repetitions. It is then recommended to control the evolution of loading more effectively and regard as valid only tests where load decreases sharply to very low levels, making sure that the three stages in the fatigue process have occurred.

1 INTRODUCTION Fatigue cracking is considered one of the most important and frequent distresses in a bituminous mixture. Among the several ways to analyse this phenomenon, three different concepts are the most usually considered: the classical criterion, the fracture mechanics based method and the energy approach. The classical criterion consists in relating the stress or strain in the bituminous mixture to the number of load repetitions to failure. Based on the traditional fatigue analysis (Whler curves), many researchers studied fatigue process through this method: Pell (1962), Pronk and Hopman (1990), Tayebali et al. (1996), Van Dijk and Visser (1977), and was taken by AASTHO as a fatigue failure criterion in 1994. The fracture mechanics approach consists in studying the crack propagation in a bituminous mixture specimen subjected to a repetitive loading. Majidzadeh et al. (1971), Roque et al. (1999, 2002), Zhang et al. (2001) used this approach to analyse fatigue process. Another way of studying fatigue failure is the damage-energy approach, which may be analysed with the constitutive damage model or the dissipated energy concept. The first alternative is based on the constitutive model developed by Kim and Little (1990), that takes into account linear viscoelasticity, healing due to a rest period and time dependence of the material. Van Dijk (1975) and Carpenter (1997) used the dissipated energy concept that defines fatigue life as a function of the dissipated energy accumulated on each loading cycle. Derived from this approach, the Dissipated Energy Ratio (DER) concept was developed by Ghuzlan and Carpenter (2000) and also used by many other researchers, such as Khalid et al. (2005).

Other researchers, Breysse et al. (2003, 2004) have also studied and modelled the influence of rest time on damage during fatigue tests and have shown the potential precariousness of the healing. The fatigue failure process of bituminous mixes has been studied in the Road Research Laboratory of the Technical University of Catalonia through the classical theory of fatigue failure and special attention has been paid to the strain evolution in the fatigue fracture region. Flexural and direct tensile fatigue tests were conducted to determine the strain evolution during the fatigue process of a series of bituminous mixtures with the aim of determining whether there is or not a certain level of permanent strain at which the mix fails due to fatigue process, irrespective of the stress or strain level applied. 2 MATERIALS AND METHODS Different semidense bituminous mixtures with a maximum aggregate size of 20 mm have been tested through three and four point bending beam tests and direct tensile fatigue test. Although they have been prepared with different RAP contents and different bitumen types, the grading was kept constant (Table 1), as well as total bitumen content (4.5% by mass of aggregate). Mixtures with 30%RAP used 80/100 penetration bitumen, mixtures with 50%RAP used 150/200 penetration bitumen and mixture without RAP were fabricated with 60/70 penetration bitumen. The average grading of the aggregates is summarised in Table 1. Table 1. Average grading for the bituminous mixtures studied
Sieve Size (mm) 25 20 12.5 8 4 2 0.5 0.25 0.125 0.063 Passing (%) 100 91.8 71.3 59.6 41.8 28.8 14.4 10.3 7.1 5.3

Two flexural fatigue tests have been used: four and three-point bending beam tests (the last one standardized in Spain), and also a dynamic direct tensile test developed in the Road Research Laboratory of the Technical University of Catalonia, with the objective of comparing the results obtained with the bending beam tests and another fatigue test type. The tests have been carried out at different temperatures: 5, 20 and 35C, in order to analyse different mixture performances: stiff and fragile behaviour at low temperatures, flexible and ductile behaviour at high temperatures. Four point bending beam was carried out according to UNE-EN 12697-24, Annex D. Figure 1 shows the four point bending equipment, composed of two inner and two outer clamps symmetrically placed. The two outer clamps keep the beam fixed and the two inner clamps are loaded to create a constant moment.

Figure 1. Four point bending beam test. Three point bending beam was carried out according to Spanish Standard NLT-350 (or UNEEN 12697-24, Annex C), with a prismatic specimen laid on its ends and fixed in its centre. Figure 2 shows the anchoring devices for the specimen testing.

Figure 2. Anchoring devices for the specimen testing. The dynamic bending beam tests (both three and four point beam tests) consist in subjecting a prismatic specimen to a time-variable displacement according to the following law:

D = D0 .sin( 2ft )

(1)

where D = displacement at moment t; 2D0 = total amplitude of the displacement function; f = wave frequency; and t = time.

Direct tensile test was developed at the Road Research Laboratory of the Technical University of Catalonia and consists in subjecting a prismatic specimen to tensile stress. Prismatic specimens can be obtained from laboratory-made prismatic or cylindrical specimens of different sizes depending on the compaction equipment used. They must be properly sawn to achieve approximate dimensions of 150x50x50 mm. Alternatively, they can be obtained by sawing cylindrical cores extracted directly from the pavement layer. In both cases, a small 5 mm indentation is made at both sides of the central section of the specimen. A metallic support is stuck on the bases of the specimen so that clamps placed on each of the press pistons are fixed to the specimen. In this way, tensile stress can be applied to the specimen, Figure 3a. During the specimen testing, the variation of strain produced in the mixture is recorded with one or two extensometers placed on one or both indented sides of the specimen, respectively, Figure 3b.
Fixed base

Extensometers Bituminous mixture specimen

Piston movement, static mode

Press piston

(a) Figure 3. Direct Tensile Test, (a) photo, (b) scheme.

(b)

Under stress controlled mode, tests were carried out at a frequency of 10 Hz and a cyclic loading is applied according to a sinusoidal function expressed in equation (2). During the test, strain evolution with the load applications is registered.

F = F0 .sin( 2ft ) + A 2 F0 = Fmax Fmin A = Fmin + ( Fmax Fmin ) / 2

(2) (3) (4)

where F = load at moment t; 2F0 = total amplitude of the load function; Fmax= maximum load; Fmin= minimum load; f = wave frequency; and t = time. 3. RESULTS AND DISCUSSION When the beam is subjected to a four point bending test, its modulus decreases with the increase of the number of cycles applied. The value of the initial modulus is calculated from the measured values of force, displacement and phase lag after the hundredth cycle. According to the classical fatigue failure criterion, the fatigue test continues until the modulus drops to half its initial value or until the specimen breaks. However, if the test is stopped at that moment of 50% modulus reduction (and the specimen is not broken), and the strain amplitude is increased, the specimen will behave as if it is not broken, since the modulus turns to be high and the test will continue until the modulus reaches half its initial value. Figure 4 shows the load evolution,

directly related to modulus evolution, where the specimen was tested according to the above mentioned procedure.
250

200

Load (N)

150

100

50

0 0 200000 400000 600000 800000 1000000

Number of Cycles

Figure 4. Load evolution with the number of cycles. Four-point Bending Beam Test at 20C, semidense bituminous mixture. So, the authors put forward that a mistake can be made if the specimen failure is considered when its modulus is reduced to a 50% of its initial value and the strain evolution is not taken into account, since the specimen may be considered to have failed much before it actually has. Flexural bending beam tests present this problem when they are performed at displacementcontrolled mode, especially when deformable mixtures containing polymer-modified bitumens or high bitumen contents are tested. In order to show this fact, and considering that the four point bending beam test does not allow registering the strain evolution at the bottom of the specimen, the authors use the three point bending beam test, where an extensometer can be fixed to the face of the beam, Figure 5.

Figure 5. Three-point Bending Beam Test. Extensometer placed to measure the strain evolution. The results obtained from the test conducted in the dynamic mode reveal that, for each mixture type, there is a strain from which the fatigue process proceeds very rapidly. This strain, which we have called critical strain, is independent of the stress state to which specimens are subjected during the fatigue process. That is, if a high stress is applied, the initial strain will be greater and will increase with each load application until the critical strain level is reached. At this point, the fatigue process will speed up and the strain level produced in each cycle will increase until the material cracks. The initial strain will be lower if a smaller load is applied,

but it will increase with each load application until reaching the critical strain and the final situation will be similar. Figures 6 and 7 show some of the results obtained with the three-point bending beam, including the critical strain level for each test, Alonso (2006) and Rodrguez (2009). Each specimen was tested at the displacement amplitude indicated in the legend of the figures. Critical strain is a parameter which depends on the type of test and tested mixture. Although test results exhibit some dispersion, a feature common to all fatigue tests, critical strain is noticeable.
0,0014 Specimen 1 (260) 0,0012 0,0010 Specimen 2 (190) Specimen 3 (200) Specimen 4 (240) Specimen 5 (280) 0,0008 0,0006 0,0004 0,0002 0,0000 0 50000 100000 150000 200000 250000 300000 350000 400000 Critical Strain Specimen 6 (280)

Strain (mm/mm)

Number of Cycles

Figure 6. Strain evolution with the number of load cycles. Three-point Bending Beam Test at 5C, semidense bituminous mixture with 30% RAP.

0.00150

0.00125 Specimen 1 (200) Specimen 2 (190) Specimen 3 (220) Specimen 4 (210) 0.00075 Specimen 5 (230) Specimen 6 (230) 0.00050
Critical Strain

Strain (mm/mm)

0.00100

0.00025

0.00000 0 500000 1000000 1500000 2000000 2500000 3000000

Number of Cycles

Figure 7. Strain evolution with the number of load cycles. Three-point Bending Beam Test at 5C, semidense bituminous mixture with 50% RAP.

Therefore, fatigue critical strain may be regarded as a mixture characteristic irrespective of the stress state to which specimens are subjected in the test. Three-point bending beam tests have also shown that mixtures not reaching the critical strain level, Figure 8, are not broken despite having attained the level of conventional failure, i.e. a 50% reduction of the load applied at the beginning of the test, cycle 200, Figure 9. The specimen is highly deformable because the test temperature is 35C. If the test results for strains and stresses are analyzed, it is observed that the initial load is reduced to half after 200,000 applications. According to the classical criterion, the specimen would have failed. However, observation of the strain evolution shows that it does not increase but rather remains stable. It would appear that the outer fibres have cracked whereas the inner fibres remain intact since the displacement necessary to make them crack is not significant enough, Figure 10.
0.0050 0.0045 0.0040

displacement amplitudes: 200 and 500 microstrain

Strain (mm/mm)

0.0035 0.0030 0.0025 0.0020 0.0015 0.0010 0.0005 0.0000 0 100000 200000 300000 400000 500000

Number of Cycles

Figure 8. Strain evolution with the number of cycles. Three-point Bending Beam Test at 35C.
600 500 400

displacement amplitudes: 200 and 500 microstrain

Load (N)

300 200 100 0 0 100000 200000 300000 400000 500000

Number of Cycles

Figure 9. Load evolution with the number of cycles. Three-point Bending Beam Test at 35C.

Kim et al. (2006) found a transition point for asphalt matrix mixtures with a new type of fatigue testing apparatus when assessing the decrease of dynamic modulus with the number of loading cycles. Two different rates of change in stiffness were observed, which were possibly indicating the limit between microcracking and macrocracking. So, this point was proposed as a failure criterion by the authors, Kim et al. (2008). It is possible that critical strain indicates the same situation, although more analysis should be done to confirm this hypothesis.

Load Non-cracked Zone

Crack Depth Bituminous Mixtures

Figure 10. Three-point Bending Beam Test. This critical strain in fatigue tests can also be observed in direct tensile fatigue tests under controlled stress, Figure 11.

0.00070
Specimen 1 (10-200) Specimen 2 (10-300)

Max Strain - Min Strain (mm/mm)

0.00060

Specimen 3 (10-250) Specimen 4 (10-280)

0.00050

Specimen 5 (10-220) Specimen 6 (10-170) Specimen 7 (10-200)

0.00040

Specimen 8 (10-180)

0.00030

0.00020

0.0001 0

Critical Strain

0.00000 0 200000 400000 600000 800000 1 000000 1 200000

Number of Cycles

Figure 11. Strain evolution with the number of load cycles. Direct Tensile Test at 5 C, semidense bituminous mixture with 30% RAP. Under controlled strain, this type of testing shows that, if the strain level is low, the specimen behaves almost elastically and the necessary load for each cycle varies very little. If the strain level is increased, the process is similar until there is a strain value under which load decreases quickly and fatigue failure takes place, Figure 12.

240 200 160 120 80

0.00005 to 0.0002

Strain Amplitude Ranges

0.00005 to 0.0003 0.00005 to 0.0004 0.00005 to 0.0005 0.00005 to 0.0006 0.00005 to 0.0007

Load (kg)

40 0 -40 -80 -120 -160 -200 -240 0 100000 200000 300000 400000 500000 600000 700000

Number of Cycles

Figura 12. Direct Tensile Test under controlled strain at 20C. Maximum and minimum load evolution with the number of cycles. Fatigue critical strain is a mixture constant which varies with mixture type and test temperature. Moreover, the analysis of the strain evolution with the number of cycles shows that the higher the mixture modulus, the smaller the increase in strain for each load application, Figure 13. It seems that critical strain and mixture modulus determine the fatigue behaviour of the mixture. It can therefore be deduced that the higher the mixture modulus and critical strain, the better its fatigue behaviour.

Mixture B Critical Strain for mixes B and D Critical Strain for mixes A and C A and B: Low Modulus Mixtures Mixture A Mixture C Mixture D

C and D: High Modulus Mixtures

Figure 13. Influence of modulus and critical strain on fatigue failure. The fatigue law is expressed by the following equation:

= a.N b

(5)

where N = number of load applications to failure; = strain; and a y b are experimentally determined coefficients. Table 2 shows the modulus, the critical strain and the fatigue laws obtained for the studied mixtures considering the classical criterion and the proposed critical strain criterion for the

extreme temperatures studied: 5 and 35C. Critical strain was determined with the bisecting line of the tangents to the strain evolution curve where the slope changes remarkably. Table 2. Modulus, critical strain and fatigue laws for mixtures with 30%RAP tested at 5 and 35C.
Temperature Testing (C) 5 35 Modulus (MPa) 16616 2007 Critical Strain (mm/mm) 0.0004 0.0016 Fatigue Law Classical Criterion = 0,0007N-0,1327 = 0,0008N-0,0874 Fatigue Law Critical Strain Criterion = 0,0006N-0,1265 = 0,0044N-0,2052

Fatigue laws at 5C are very similar, but the results obtained at 35C show higher fatigue life when it is calculated with the Critical Strain criterion (this life comparison is valid for a number of cycles less than 1,000,000), Figure 14.
0.01 Classical Criterion 5C Critical Strain Criterion 5C Classical Criterion 35C Critical Strain Criterion 35C

Strain (mm/mm)

0.001

0.0001

0.00001 1000

10000

100000

1000000

10000000

Number of Cycles

Figure 14. Fatigue laws for the mixtures with 30%RAP tested at 5 and 35C. 4. CONCLUSIONS The results of this study reveal that fatigue cracking of mixtures takes place when a certain strain level is reached, here named fatigue critical strain, for each test type and temperature, independently of the applied stress. Mixtures with the same critical strain level but different modulus exhibit different fatigue behaviour. It can therefore be deduced that the higher the mixture modulus and critical strain, the better its fatigue behaviour. The results have also shown that, when performed at controlled strain rate, flexural bending beam tests may yield false results when highly deformable mixtures are tested, as is the case of mixtures prepared with polymer-modified bitumens. 5. REFERENCES

Pell, P.S., 1962, Fatigue characteristics of bitumen and bituminous mixes. In Proceedings of the 1st International Conference on the Structural Design of asphalt pavements, Ann Arbor, Michigan, 20-24 August 1962. Department of Civil Engineering, University of Michigan, Ann Arbor, Mich., pp- 310-323. Pronk, A. and C. Hopman, 1990. Energy Dissipation: the Leading factor of fatigue. Proc., SHRP Conference Highway Research: Sharing the benefits, London, United Kingdom, 1990, pp. 255-267. Tayebali, A.; J. Deacon and C. Monismith, 1996. Development and evaluation of dynamic flexural beam fatigue test system. In Transportation Research Record 1545, TRB, National Research Council, Washington, D.C., 1996, pp. 89-97. Van Dijk, W. and W. Wiser, 1977. The Energy Approach to fatigue for pavement design. In Journal of the Association of Asphalt Paving Technologists, Vol. 46, 1977, pp. 1-40. Majidzadeh, K., E.M. Kauffmann and C.L. Saraf, 1971. Analysis of fatigue of paving mixtures from the fracture mechanics viewpoint, In ASTM Spec Tech Publ 508 (1971), pp. 6784. Roque, R.; Z. Zhang and B. Sankar, 1999. Determination of crack growth rate parameters of asphalt mixtures using the Superpave IDT, In Journal of the Association of Asphalt Paving Technologists 68 (1999), pp. 404433. Roque, R., B. Birgisson, B. Sangpetngam and Z. Zhang, 2001. Hot mix asphalt fracture mechanics: a fundamental crack growth law for asphalt mixtures, In Journal of the Association of Asphalt Paving Technologists 71 (2002), pp. 816827. Zhang, Z.; R. Roque, B. Birgisson and B. Sangpetngam, 2001. Identification and verification of a suitable crack growth law, In Journal of the Association of Asphalt Paving Technologists 70 (2001), pp. 206241. Van Dijk, W., 1975. Practical fatigue characterization of bituminous mixes. In Journal of the Association of Asphalt Paving Technologists 44 (1975). pp 38-74. Carpenter, S. and M. Jansen, 1997. Fatigue behaviour under new aircraft loading conditions. In Aircraft /pavement Technology: in the midst of change, Seattle, Washington, 17-20 August 1997, edited by F.V. Hermann, American Society of Civil engineers, New York, pp 259-271. Kim, R. and D. Little, 1990. One-dimensional constitutive modelling of asphalt concrete. In Journal of Engineering Mechanics, 116 (4), pp. 751-772. Ghuzlan, K., and S. Carpenter, 2000. Energy-derived, damage-based failure criterion for fatigue testing. In Transportation Research Record 1723. Transportation Research Board. National Research Council (US) Washington, D.C. pp 141-149. Khalid, H. and I. Artamendi, 2005. Characterization of fatigue damage for paving asphaltic materials. In Fatigue & Fracture of Engineering Materials & Structures, Blackwell Publishing Inc., Sheffield, UK, No. 28, 2005, p. 1113-1118. Breysse, D.; C. De La Roche, V. Domec and J.J. Chauvin, 2003. Influence of rest time on recovery and damage during fatigue testes on bituminous composites. In Materials and Structures, Vol. 36, (2003), pp. 648-651. Breysse, D.; V. Domec, S. Yotte and C. De La Roche, 2004. Better assessment of biutminous materials lifetime accounting for the influence of rest periods. In Cracking in Pavements, RILEM Proceedings, France, May 5-8, 2004, pp. 123-130. Alonso Mota, J., 2006, Estudio del proceso de deformacin y agrietamiento por fatiga de mezclas bituminosas sometidas a carga cclica. Ph. D. thesis, School of Civil Engineering of Barcelona, Technical University of Catalonia, Barcelona. Rodrguez Cambeiro, M., 2009. Anlisis de la propagacin de fisuras y de los criterios de fallo en el comportamiento a fatiga de las mezclas bituminosas. Ph. D. thesis, School of Civil Engineering of Barcelona, Technical University of Catalonia, Barcelona. Kim, Y.; H.J. Lee, D.N. Little and Y.R. Kim, 2006. A simple testing method to evaluate fatigue fracture and damage performance of asphalt mixtures. In Journal of the Association of Asphalt Paving Technologists 75 (2006), pp. 755788. Kim, Y.; D.N. Little and R.L. Lytton, 2002. Use of dynamic mechanical analysis (DMA) to evaluate the fatigue and healing potential of asphalt binders in sand asphalt mixtures. In Journal of the Association of Asphalt Paving Technologists 71 (2002), pp. 176206.